Documentation

Up and running on the first launch.

Six steps take you from download to your first searchable clip description — about five minutes, most of it waiting on the model.

1 · Buy a license 2 · Install 3 · Approve first launch 4 · Add your AI key 5 · Activate 6 · First clip
Before you begin
A Mac with Apple Silicon running macOS 13 Ventura or newer.
An API key from Anthropic or OpenAI — or a local model server (Ollama, LM Studio) if you'd rather keep frames entirely offline.
About five minutes.
01

Buy a license

videoscriber has no free tier — the app unlocks with a one-time $79 license purchased through Gumroad. After checkout, Gumroad shows (and emails) your license key. Keep two things handy for step 5:

The license key from your Gumroad receipt.
The email address you used at checkout — the license is bound to it, and activation asks for both.
02

Download and install

  1. Download the current release: videoscriber.io/download
  2. Double-click the downloaded .zip to unpack it.
  3. Drag videoscriber.app into your Applications folder.
03

Approve the first launch

videoscriber ships notarization-free, so macOS asks you to approve it once. This is expected — here's the dance:

  1. Double-click videoscriber.app. macOS will say it "cannot be opened." Dismiss the dialog.
  2. Open System Settings → Privacy & Security and scroll down to the Security section.
  3. Click "Open Anyway" next to videoscriber, then confirm.
Terminal alternative — one line, same effect:
xattr -dr com.apple.quarantine /Applications/videoscriber.app

You only do this once. Automatic updates install normally afterward — they're delivered in-app and cryptographically verified, so Gatekeeper never asks again.

04

Add your AI key in the welcome flow

On a fresh install, videoscriber opens with a short welcome flow. It asks for one thing that matters: the API key of the vision model that will describe your footage. Frames go from your Mac straight to your provider — never through our servers.

  1. Click Continue past the welcome screen.
  2. Pick a provider — Anthropic Claude or OpenAI.
  3. Paste your API key (create one at console.anthropic.com or platform.openai.com) and click Continue.
  4. The last pane previews license activation — click Get Started to enter the app.

Prefer a local model, OpenRouter, MiniMax, or another endpoint? Finish the welcome flow, then point Settings → API at any Anthropic- or OpenAI-compatible base URL — e.g. http://localhost:11434 for Ollama. Local servers need no key at all.

05

Activate your license

The main window opens showing a 🔒 Not licensed badge until you activate:

  1. Open videoscriber → Settings… from the menu bar (or press ⌘ ,) and choose the License tab.
  2. Paste your Gumroad license key and the email you used at checkout.
  3. Click Activate. videoscriber verifies the key with Gumroad, and the badge turns green with your email.

One license covers every Mac you own — repeat this step on each machine. Activation needs an internet connection; after that, the license is re-checked quietly in the background.

06

Describe your first clip

  1. Drop a video anywhere in the window (or press ⌘ O to pick files). The clip joins the queue and starts immediately.
  2. Watch the job move through extracting → describing → writing. When it's done, two sidecar files appear next to your video: .description.json and .description.md.
  3. Switch to Library (⌘ 2), open the folder you just described, and search it by what's in the frame.

That's the whole loop. From here, drop 500 clips and walk away — the queue survives restarts — or point Settings → Watch at an ingest folder so new footage describes itself.

Good to know
.mkv and .webm files need ffmpeg. Every other common container decodes natively. If you work with Matroska files, install it once with brew install ffmpeg.
A job failed with a provider error? Check the key and model in Settings → API, then right-click the job and choose Retry. Failures never touch your video files.
There's a CLI too. The same license unlocks the videoscriber command-line tool for scripted and headless batches.
Stuck? Email [email protected] — include the job's error text if there is one.